1
\$\begingroup\$

From what I understand, RF- and microwave-circuits needs separate ground planes to avoid coupling between signal and DC/digital currents. I'm currently involved in an RF-project where we're going to use a SPDT-switch to control the path of the signal. The switch is Mini-Circuits SWMA-2-50DR+.

What confuses me is the suggested PCB layout, which includes signal-ground and TTL-ground terminals on a shared plane that's grounded through a lot of vias. I can't see how the currents can be uncoupled with this geometry, and where the vias are supposed to go if I have two ground planes. I know that separate ground planes need interconnection to tame unwanted modes, but I find it strange that it happens on the signal/component-layer.

I'm sure the layout is sensible, but I just don't have the experience to understand it yet. Enlightenment would be very much appreciated.

Best regards -paw

\$\endgroup\$
1
\$\begingroup\$

The datasheet does not make any claim that the TTL driver is isolated from the RF side of the switch, so even if you isolated the two grounds in your layout, the chip may internally tie the grounds together.

I see two options:

  1. Use an opto-coupler before the TTL input so that you are sure your digital circutis are isolated, then use the recommended layout.
  2. Buy a chip, and make some measurements verify that (a) the TTL ground is not internally connected to the RF ground and (b) the chip will still operate if the two ground planes potentials drift apart a little bit. If you are sure of both, then you can modify the layout.
\$\endgroup\$
  • \$\begingroup\$ I already have a few of these. I'll look into option 2 and measure it. Pretty obvious, but I didn't think of it myself. Thanks! \$\endgroup\$ – paw Apr 27 '15 at 21:45
  • \$\begingroup\$ I just measured the ground planes, and they are not interconnected. I think I'll separate pin 3 (TTL ground) from the shared ground plane on top and route it to the DC/digital plane. \$\endgroup\$ – paw Apr 28 '15 at 10:29
0
\$\begingroup\$

I would suggest following their layout. In general, unless your design is very noise sensitive (contains low amplitude signals or high-gain amplifiers), it will be best to simply have one ground plane and connect everything to it with the lowest possible impedance. The moment you have separate grounds, you have the potential to accidentally design a patch antenna which will cause unwanted radiation. This can become a bit of a religious issue, but this is what I have generally found. I am not a low-noise specialist, however.

Noise gets onto ground planes when large currents flow and create a voltage gradient on the plane. Digital circuits can do this due to switching transients. So do a good job bypassing your high-speed digital circuits with capacitors, and if possible, route your RF signals away from the high speed digital stuff (including high speed data lines), and you will likely be OK.

If the RF stuff is low amplitude, it would be best to insert a shielded pre-amplifier to boost the signal before the signal gets to your digital board. Then any noise added by your board will not hurt the S/N ratio much.

Another option is to put the RF switch on a separate board by itself, or in a metal can, and only route the control signals to the switch from your digital board. The control signals could be heavily filtered (for example with a passt-through ferrite bead on the cable).

Hope this helps.

\$\endgroup\$
  • \$\begingroup\$ The design is noise sensitive and I can't add a pre-amplifier because of time constraints. I have decided to try to use two separate ground planes (via interconnected) and split the common ground plane on the component layer. I'll see what happens. Thanks! \$\endgroup\$ – paw Apr 28 '15 at 12:39

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.